Technology in the foreign exchange markets has become an ever-present part of the industry in recent years, from pre-trade, electronic trading and post-trade elements all factoring into the entire trade lifecycle for asset management firms. Some of this change has been natural market progression. The market has seen a prevalent take-up in electronic trading among the buy side since the development of request-for-quote, with multi-dealer platforms and bank-owned platforms vying for market share among their clients.
In the dealer-to-customer segment of the market, electronic trading has been slowly edging upwards over the years. According to the Bank for International Settlements, the share of electronic trading has ticked up from 56% of turnover in 2013 to above 63% last year.
That has been quicker depending on the product. For example, the spot market has registered quicker growth, now sitting at 75%, and FX forwards at 66%. The swaps market has lagged and flat-lined around 55% of the market, but it is still growing.
The growth in electronic trading among buy-side firms has also facilitated new tools for market participants - taking inspiration from other markets that have been electronic for longer - to help them better understand how their trades impact the market, giving vital information into how and when is best to trade in the future.
Some of this has become mandated by regulation. While spot FX may have been left off the list of mandated products in the second iteration of the Markets in Financial Instruments Directive (MiFID II), derivatives are required to be measured for best execution by buy side firms in order that they can be transparent to clients they are trading in the most efficient way possible.
Defining how exactly to achieve best execution does not illicit just one response. One important aspect is that this is something that the buy side needs to measure themselves and not rely on the banks for.
“There are a variety of actions the buy side should take to achieve better FX execution outcomes, but the main principle should be - it is your responsibility – solely. Consistent improvement in FX execution outcomes cannot be outsourced. The buy side should know by now that It certainly is not the banks’ responsibility,” says James Singleton, chairman and chief executive at Curex.
“And while some TCA [transaction cost analysis] providers are providing better post trade insights, only buy side traders can commit to use the data provided in that analysis to improve their trading approach and outcomes. It is the buy side’s ultimate responsibility to make sure they have the right tools and access to the right pre- and post-trade data to improve their execution outcomes in real time,” he adds.
Putting prices in competition would often be the first and foremost principle in ensuring better execution and this is true up to a certain point, but the question remains - how do participants determine execution outcomes?
For some, execution outcomes can only be fully understood when they are measured against independent data that comes from outside the system that is being measured. Aside from being a regulatory obligation, independent data is seen as essential to determine the full extent of execution costs, and is thought to be worthless without it.
“To understand what is being paid, it is important to understand that FX costs consist of two parts: the clearing price and the idiosyncratic price,” says Xavier Porterfield, head of research at New Change FX. “The clearing price is the prevailing mid-rate when a customer price is set. This can be thought of as the raw material cost. The idiosyncratic price is the user specific price that is then shown to the client. This is the bespoke manufacturing cost.”
“This user specific price must be set by the broker so that the idiosyncratic cost to the client is higher than the broker’s clearing cost. Liquidity is fragmented across different venues and providers so no single source data can accurately approximate the clearing cost, and without knowing the clearing cost it is not possible to correctly identify idiosyncratic costs. Without independent data any conclusion is built on sand,” he says.
Another fundamental principle for best execution technology providers is that best execution should be defined for that firm’s specific business, clients and style of execution, rather than a ‘one size fits’ all approach to best execution.
Also, there is the need to consider what is best execution from a fiduciary perspective for the asset owners, as this may vary depending on the specific mandate in question. For example, which benchmarks are going to be selected to determine whether the execution outcome has been improved.
For Peter Eggleston, co-founder and director at BestX, once those over-arching principles are defined, there are a number of additional themes to help achieve better execution outcomes.
“Buy side firms need to ensure they have a complete, clean trade data set with accurate time stamps; use analytics that are relevant for FX, providing robust output that all stakeholders agree with; view best execution as a process, from pre-trade through to post-trade evaluation and monitoring, with a feedback loop to enable incremental improvements based on rigorous, data driven analysis. And finally, measure and monitor all trades but only make performance improvement decisions based on meta-analysis of statistically significant sample sizes of trades. On any one trade, random market events may result in random execution outcomes that do not necessarily mean that best execution wasn’t achieved,” he says.
For algorithmic execution, buy side firms will typically find that algo orders lend themselves to be easily reviewed and studied since all the order information can be databased and brought to life through TCA.
“We continue to hear that buy-side traders are having success improving their execution quality by leveraging execution algos. Part of the success in this experience is the underlying benefit execution algos provide – like anonymity, faster decision making and implementation on order routing than humanly possible, reduced market impact but another part of the success for execution algos is the ability for the buy-side to easily study and review past trades through TCA in order to make improvements in the algo behaviour for future trades,” says Curtis Pfeiffer, chief business officer at Pragma.
As the FX market has evolved and become more electronic, buy side firms have been able to be more creative in reducing market impact from trades with the advent of algorithmic trading. Achieving minimal market impact is one of the ultimate goals of best execution and is viewed as a touch challenge for the buy side, and one algos can potentially help with.
Market impact is the cost incurred from participating in the market, and in the past a trade could be done in one fell swoop. When an order is made for the full amount, market impact is a one off cost that is manifested in the difference between the clearing cost prevailing in the market when the order was given and the price where the order was filled. Now, with more electronic trading, algos have been seen as the answer in order to minimise the impact of trading for buy side firms.
Algos also come in different shapes and sizes. Buy side firms need to understand how intelligent the algo should be, whether it should be aggressive or passive, how fast it should execute, and what kind of liquidity should the algo be willing to interact with. Different liquidity pools and how they operate - whether they have last look or not - and the types of participants in those pools - such as high frequency traders - can also affect market impact.
“A very important component is to review executions and capabilities of providers on a consistent basis and to ensure one has enough data before making modifications. In algorithmic trading, determining the signal from the noise can sometimes be challenging but it is critical in order to demonstrably improve execution quality,” says Pfeiffer.
“Also critical is the realization that when the buy side uses algos, they retain market risk, so complete anonymity during the execution is paramount for best execution and low market impact,” says Singleton.
For an algorithm with a pattern of repeat orders, market impact costs are cumulative. Instantaneous spreads at the moment of execution can be set to appear attractive, but the value to the dealer lies in knowing a client’s entire order, so how a dealer clears the trade after each fill will determine whether subsequent trades are relatively more or less expensive. So executing a large order in multiple slices can incur large costs - both by prices moving against a client during the trade or by prices moving less favourably in the client’s direction.
For example, a buy order in a falling market momentarily supports the market, and the price would have been lower, absent the buy order. Correctly measuring market impact allows clients to see how, where and with whom price movement materially affects execution outcome.
It stands to reason that selecting the dealers with which a firm chooses to trade is of paramount importance when trying to achieve best execution, but that is not possible without an understanding of market impact costs, which can reveal whether a dealer is simply funnelling orders through the market or tactically managing the client’s risk.
“Reducing market impact costs is a question of selecting the best dealer with the best algo for a given pair in given conditions. The problem is that market impact costs are circumstantial so understanding how well an algo actually performs requires an understanding of its unit cost of volatility. Part of transaction costs are driven by volatility,” says Porterfield.
“We need a way to normalise transaction costs for volatility conditions to understand how much a trade costs relative to the riskiness of a trade. Higher volatility should incur higher impact costs, to cover the inventory risk of a holding a trade. By the same measure, lower volatility should incur lower costs because the risk is lower. By measuring dealers by their unit cost of volatility we can compare dealers and trades occurring in different currencies and time periods,” he adds.
From an algo perspective, the implications of market impact can be significant in terms of the overall price achieved by the algo. For example, if an algo is buying AUDUSD over the course of an hour, and the first few child executions of the algo create significant market impact, pushing AUDUSD higher, then the remainder of the algo will be buying at raised prices, resulting in a less beneficial outcome overall.
Such impact can be created in a number of related ways, such as through an overly aggressive execution by the algo relative to the prevailing liquidity conditions, or by ‘signalling’ to the market, allowing other participants to take advantage of the buy signal.
“Analytics and data can help to mitigate market impact through appropriate parameterisation of the algo (e.g. ensuring the algo is given the required amount of time to execute the required notional without an excessive participation ratio) and also through informed decision making of algo style, specific algo type, timing etc,” says Eggleston.
In a world of more electronic execution, the splitting up of orders is viewed as a necessary step by some to achieve better execution. These child orders can help reduce the market impact, too.
“The routing of each child order involves the observation of many data points in fractions of a second which if done well, can lead to more consistent and better execution quality. It is here, where the law of large numbers often prevails over an instantaneous risk transfer, which can require a big price concession,” says Pfeiffer.
Encouragement to change
While the tools may be available to help achieve best execution, the buy side needs to adopt what is out there in the marketplace in order to fully get the most out of these now not-so-new concepts in the FX industry.
For example, according to market consultants, only half the buy side even subscribes to third-party TCA, despite there being many tools available to the buy side to use in the pre- and post-trade environment.
“The buy side will fully engage in the best execution process when they finally realize that they actually benefit from its practice and pursuit. Firms like ours are providing real-time performance measurement while our clients trade using algos. The buy side has been slow to embrace the technology that has been built to help them achieve best execution,” says Singleton.
Strangely, given the clout that many on the buy side have, is the fact that many give up control rather regularly to the sell side as clients are told by their banks what are the best tools or processes to use.
“The banks have access to much more information than the buy side so the buy side needs to equip itself with tools that level up their information deficit. Best execution describes a process that calls on fiduciaries to take all sufficient steps to ensure execution quality,” says Porterfield.
The struggle among some corners of the buy side to take up best execution tools is shown by their lack of general interest in the FX Global Code - a set of principles of good practice in the FX market, with the aim of creating a robust, fair, liquid, open, and appropriately transparent market.
(See our Market Commentary article in this edition)
“Good providers are currently mixed in with bad, but the apathy towards best execution processes means that the differentiation is not made. By engaging with the sell side and requiring additional transparency from suppliers, buy side firms can make a significant difference to their investors,” says Porterfield.
Given the squeezing of fees on the buy side and general cost pressures that the community is facing, best execution can help in that process. More efficient execution translates into higher net asset value, but the challenge is to make it easy for the buyside to lower their search costs. This is where independent benchmarks play a crucial role.
Another crucial element for adoption is education, according to Eggleston. “It can be a surprise to asset owners and managers to hear that the FX market isn’t as liquid as they may have thought, especially if they come from an equity background, and consequently just how much of a difference can be made to portfolio performance when more attention is paid to the FX best execution process,” he says.
He feels it is critical to measure execution outcomes accurately as it is impossible to manage something that cannot be measured. This includes the availability of accurate timestamps for every trade. When the accurate measurement of FX costs and performance become available at an institution, in a format that allows actions to be taken to improve outcomes, BestX generally finds that the best execution process evolves from purely ‘box-ticking’ to one where institutions become more actively engaged and look to make changes to processes and add value.
“In addition, there is an increasing commercial pressure from asset owners who are more actively requesting information from managers on their best execution processes when putting mandates out to tender. This is naturally resulting in increased engagement as it is becoming a core component of winning new business,” says Eggleston.
New age tec
As the FX market continues to become more electronic, new technologies such as artificial intelligence and machine learning will also start to play a role, say experts. Given the amount of the market that is traded electronically, some believe the data already exists to create the foundation for those advancements, however fragmentation of the market is holding up that process.
“The technology exists to execute differently so now it is purely a question of the buy side requesting the service from their suppliers,” says Porterfield. “The purpose of a benchmark is to identify the clearing cost of a trade. It is then simple to see the idiosyncratic costs that are piled on top of that clearing cost.”
One such next step could be ‘real time’ TCA, as opposed to pre or post-trade TCA. “Real Time TCA using independent benchmarks allows clients to control the execution process. When the benchmark is independent of the execution process it provides dealers the opportunity to benchmark the cover trade and charge defined fees. Fills can be tracked, with pre-agreed limits in terms of tolerable slippage to benchmark, providing an audit trail of the process, and automating TCA at source,” adds Porterfield.
Using technology, data and analytics to help traders make more informed decisions to achieve better execution outcomes is a core trend for the future. While tech will play a big role, a combined approach of using technology in conjunction with trader experience and market knowledge, will aid the decision making process.
As market structure becomes more complex, traders are being asked to trade more asset classes and products, using more protocols and on more venues, so harnessing data is going to be key to help navigate the complexity.
Low or high touch
“Designating trades as low or high touch will be critical for success as traders will increasingly need to focus their time on the more complex trades, which may require more manual intervention to achieve the optimal outcome. In order to free up the time to do this, in a world where cost cutting is rife, will most likely require the more liquid transactions to be designated as low touch, where rules based automated execution will dominate,” says Eggleston.
“However, a robust and rigorous analytics framework is required to initially define such rules, and then also to ensure that the rules are performing as expected through both tactical and strategic performance monitoring. The results of the strategic monitoring should then feedback into the rules engine to ensure it remains up to date and relevant for the market structure and liquidity conditions in which it executes,” he adds.